It's rare to find the opportunity to benefit the developers community through your job, it's even more rare when the company's leadership motivate and help you do so! We wrote the blog post as a "I wish someone told me this 5 years ago" mentality. We open sourced and currently maintain production code to show an example for the ml community how to utilize ML at scale with real life non controlled environment scenarios.

GitHub

GitHub - tidal-music/tidal-algorithmic-mixes: Tidal algorithmic mixes

Tidal algorithmic mixes. Contribute to tidal-music/tidal-algorithmic-mixes development by creating an account on GitHub.

This is how we try to utilize our passions in machine learning to help people find new music! And we'll love your help discussing, critiquing, and contributing to it! View quoted note →

Gotta love the simple things in Oslo #kolsåstoppen

I think AI will become more open sourced and more democratized, OpenAI will not continue to be the leader for the next decade, the next leader will be no one. or no single player. Here's my 2 cents on it. In my very short career as a machine learning engineer I worked with AI in an increasingly expanding scope. From a product offering standpoint, the ML field has evolved a lot. 10+ years ago we had niche features like brand health through social medial sentiment analysis or half broken OCRs. With break throughs like Alexnet, and progress through the Netflix prize, it started becoming a more important but a nice to have part of the product, recommending people you may know in social media, or movies to watch. With more advances it started becoming more integral to the core user experience, and the product success or failure, examples include social media algorithms or Tidal algorithmic mixes. This is also the time AI started to become too strong, being labeled as dangerous or unethical due to biases in data from our behavior. With the break throughs of attention is all you need, and the transformer architecture, where the authors aim of creating one model to rule them all for natural language problems, it was only a matter of short years before seeing massive breakthroughs, from Bert, to GPT and the rise of LLM. The new era of AI powered products will take it from a crucial component to a foundational one that the product is built on. I feel like there is a great parallel between AI and OS rise stories, small niches -> nice to haves -> integral parts -> foundational systems. Currently Unix based systems dominate the world, as once a technology becomes core to everyone's success, no one player can or should posses the ability to hoard that power. The future of AI is democratic, the past present and future are all open source!

Companies like the hugging face 🤗 give me hope that closed source AI movements are not the future, the past, present and future is open source, IMO computer science is built on that premise, and I truly hope closed source knowledge hoarders change their eays.

#bigcode #chatgpt #copilot #opensourcesoftware #github #developers #machinelearning #deeplearning | Philipp Schmid | 119 comments

Introducing StarCoder ⭐️, the biggest open-source Code-LLM 🤖. StarCoder is a 15B LLM trained on permissively licensed data from GitHub. It o...

So it's 25th of April and it's snowing in Oslo

Sup nostr world!